Why did Jensen Huang come back? Understanding his visit is key to spotting the next investment opportunity

Nvidia CEO Jensen Huang returned to South Korea just eight months after his so-called "kkanbuchicken meeting" sent shockwaves through the domestic market last October. This visit was no mere ceremonial call. The itinerary stretched across four nights and five days, and Huang held a series of back-to-back meetings with the heads of South Korea's major conglomerates — signaling that the trip carried real strategic weight. The visit offered a telling glimpse into how Nvidia, the central player in the global AI industry, views South Korea, and what shifts in the industrial landscape Korean companies may soon face.

Markets have largely interpreted the visit as a deepening of the AI alliance between Nvidia and South Korea's top conglomerates. But it is worth looking beneath the surface. In the past, Korean companies were seen primarily as major consumers of high-performance semiconductors or, in some cases, as component suppliers. That dynamic is changing. As the AI industry evolves beyond a GPU sales race into a comprehensive infrastructure sector — one that integrates data centers, power, cooling, communications, servers and operational technology — the role Korean companies can play is expanding accordingly.

The core message of Huang's visit is clear: Nvidia sees South Korea not merely as a customer or a link in the supply chain, but as a strategic partner in designing the AI era ahead.

The substantive focus of the visit appeared to rest more heavily on the AI data center business than on robotics. Robotics represents an enormous long-term market, but meaningful revenue visibility is still some way off. AI data centers, by contrast, are already the site of a global investment race — and the area where Nvidia's interests and those of Korean companies align most quickly.

A significant share of Nvidia's current revenue comes from global hyperscalers. Microsoft, Amazon, Google and Meta — the giants of cloud computing — are the primary demand drivers of the generative AI era. They have been buying Nvidia GPUs in bulk and leading the charge on AI infrastructure. At the same time, all of them are accelerating the development of in-house chips to reduce their dependence on Nvidia. The relationship between Nvidia and the hyperscalers is closer to an adversarial symbiosis — cooperation and competition coexisting.

The problem is that capex growth among these hyperscalers will inevitably slow. AI infrastructure investment cannot keep expanding at its current exponential pace. Global big-tech companies are already carrying enormous data center investment burdens, and the challenges of securing power, arranging financing and navigating regulation are growing more pressing. For Nvidia, diversifying its revenue base is a necessity. Expanding into sovereign AI, enterprise AI and industrial AI infrastructure markets is less a choice than an inevitability.

South Korea is emerging as a prime candidate for that role. It is one of the rare countries that possesses the full stack: semiconductor manufacturing, memory chips, telecommunications, power equipment, construction and data center operating experience. Building an AI data center is not simply a matter of stacking GPUs. It demands a complex set of capabilities — massive power intake, stable cooling, high-speed networking, server manufacturing, memory supply, operational software, site acquisition and permitting. On all of these dimensions, South Korea meets much of what Nvidia needs in an AI infrastructure ally.

The challenges, however, are considerable. Hyperscale AI data centers require power measured in gigawatts, and the capital outlays are astronomical. Site acquisition, permitting, grid interconnection, power intake capacity, cooling infrastructure and community acceptance all represent significant hurdles. Ultimately, the race to build AI data centers is simultaneously a semiconductor race and an energy race. The bottlenecks in the AI industry are likely to shift — from GPU shortages to power shortages, transmission grid constraints and cooling infrastructure deficits.

That is precisely why Naver Cloud, SK Telecom and other domestic AI and cloud companies drew attention during this visit. Naver has been among the most concrete in articulating a sovereign AI strategy within South Korea. It has a blueprint for building a Korean AI ecosystem anchored in its own large-scale AI models and cloud infrastructure. Sustaining large-scale data center investment, however, will require substantial financial firepower and the ability to secure reliable power.

SK Group brings a different set of strengths. SK hynix holds a pivotal position in the global AI semiconductor supply chain, driven by its leadership in high-bandwidth memory. The group can also draw on affiliates spanning telecommunications, energy and infrastructure — SK Telecom, SK Broadband and SK eternix. The ability to recycle cash flows from memory chips into data centers and AI infrastructure gives SK Group a relatively well-prepared business structure for what lies ahead.

This is also why Huang's visit carries particular significance. A central question now preoccupying the market is where the enormous cash flows that Samsung Electronics and SK hynix stand to generate from the AI semiconductor cycle will ultimately be reinvested. In the past, chipmakers directed their capital primarily toward expanding production capacity and advancing process technology. In the AI era, however, the investment frontier can extend well beyond semiconductor manufacturing — into data center operations, power infrastructure, server systems and AI platforms.

SK Group is moving in that direction with relative speed. It appears to be positioning the data center industry as the next stage beyond its semiconductor business, with group-wide attention expanding accordingly. This can be read as a strategic shift that goes beyond simple supply chain participation — one that encompasses both manufacturing and operations. Where Korean companies once played the role of component suppliers or manufacturing partners in the global IT industry, the possibility of entering the AI infrastructure business as direct operators is growing.

From an investment standpoint, energy is the first variable to watch. AI data centers are extraordinarily power-intensive. As GPU performance rises, so does power consumption. The growth of the data center industry will therefore likely drive broad demand across the energy infrastructure spectrum — power equipment, energy storage systems, fuel cells, cooling systems, power conversion devices and distribution equipment. Related thematic stocks may attract attention in the near term, but over the longer run the key is identifying companies that have secured actual revenue from global customers.

Against that backdrop, companies in the data center infrastructure space are likely to become a major market focus. Those with a high proportion of overseas revenue and a track record of validation from global customers stand to command a premium. AI infrastructure investment is not confined to any single country — it is unfolding simultaneously across the United States, the Middle East, Europe and Asia. A Korean company that has secured a place in the global data center supply chain has a far larger growth opportunity as a result.

The semiconductor equipment sector also warrants renewed attention. Capital expenditure is likely to expand in the second half of this year, centered on memory chips and HBM. The fact that Nvidia is pressing Korean semiconductor companies for greater production capacity is an important market signal. With structural growth in AI semiconductor demand, a capex ramp-up at Samsung Electronics and SK hynix may be only a matter of time.

That said, semiconductor equipment stocks always carry valuation risk at the early stages of a capex cycle. Share prices and expectations tend to move ahead of actual earnings confirmation. Rather than approaching the sector purely on the basis of a recovery in industry conditions, investors should also examine new equipment development, market share gains within key customers and the potential for margin improvement. Some equipment companies have recently shown more stable profitability than in the past, as their collaborative relationships with customers have deepened.

From this perspective, semiconductor equipment companies such as TES merit attention. A turnaround has been visible since 2024, and new equipment development combined with closer collaboration with major customers has raised the prospect of improving profitability. If the AI semiconductor cycle plays out around HBM and high-performance memory investment, related equipment companies stand to benefit over the medium to long term. Given that this is a sector where share prices can move ahead of earnings, balancing earnings confirmation with investment timing is essential.

On the data center infrastructure side, Seojin System is worth watching. The company operates across three business pillars — energy storage systems, solid oxide fuel cells and semiconductor equipment manufacturing — all of which connect to the broader AI data center theme. Data centers require stable power supply, backup power and energy efficiency. ESS, fuel cells and power infrastructure components all stand to play an important role as data center investment expands.

Its global customer base is a particularly important factor. In the data center infrastructure market, overseas revenue and inclusion in global supply chains carry more weight than domestic thematic exposure alone. Companies that have passed quality validation from global players are well positioned to expand their order pipeline as AI data center investment accelerates. That is the core reason Seojin System is attracting attention.

In global markets, Nvidia's next-generation AI chips and the large-scale data center infrastructure needed to run them are being discussed in the same breath. The more powerful the AI chip, the more critical the power infrastructure becomes. As the power demand of a single data center grows to the scale of a mid-sized city, the power grid, transformers, ESS, cooling systems and fuel cells are no longer peripheral industries. They are the core underlying infrastructure of the AI industry.

Ultimately, Huang's visit signals a structural shift in the areas where South Korea can lead in the AI industry. The country's economic strengths were once rooted in labor-intensive manufacturing and efficient supply chain management. That evolved into capital-intensive advanced manufacturing through semiconductors, displays and secondary batteries. Now, the possibility of a further evolution — into hyperscale technology infrastructure encompassing AI data centers and energy infrastructure — is opening up.

The market implications are also significant. There has been criticism that the recent concentration in Kospi-listed stocks has gone too far, but there are clear reasons behind it. Global capital is responding more to growth potential and return on equity than to interest rates. The market's core engine right now is the growth story of companies with strong profitability that can deploy additional capital. It resembles the moment when a thriving shop owner takes out a loan to open more locations.

That does not mean expecting a broad-based rally across all stocks. Even within the AI infrastructure theme, genuine beneficiaries must be distinguished from companies riding the narrative. Semiconductors, data centers, power, ESS, fuel cells, cooling and communications infrastructure will all be mentioned in the same breath, but what ultimately matters is customers, revenue, margins, financial strength and execution. The Kosdaq may see some recovery in the second half of the year, but selective positioning will be more important than ever.

The conclusion from this visit is clear. Competition in the AI era is no longer purely a contest of software models. Only countries and companies that can secure GPUs, process data, supply power, manage heat and operate reliably will advance to the next stage. That is why Nvidia came back to South Korea. Korea has the potential to be not merely a consumer in the AI infrastructure race, but a meaningful architect and executor.

Huang's visit should not be dismissed as a short-term share-price event. It raised a question about where Korean industry will allocate capital over the next decade. If the cash flows generated from semiconductors are reinvested into data centers and energy infrastructure, the leading industries on the Korean stock market could be reshuffled once again. At its core, AI is about computing power — and computing power is ultimately about electricity and infrastructure. That is precisely where the market should be looking now.